关于「实证闭环:删入口的整时段切换实验、组织层级审计与预注册方案」,《入口加法与协同层的均衡代价》给出的定量结论是什么?

结论:第八章说明按用户分流的实验会把增加 3.800 分钟的入口报告成缩短 2.616 分钟,因而改用整时段切换;给出蒸馏小工具「入口删减均衡计算器」与预注册方案,删入口实验每组 49 天,层级审计每组 99 个请求。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。

数据口径:书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,任何删入口、撤协同层的生产变更须由主事者人工复核放行。 书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 49, 99

z_alpha + z_beta = 2.8016
(1) days per arm for the entry switchback
deltacv 0.10cv 0.15cv 0.20
0.05063142252
0.085224987
0.100163663
0.15071628
(2) requests per arm for the layer audit
ratios 0.4s 0.6s 0.8
1.102776231106
1.2076171303
1.274499176
1.50163562
查看完整验算程序
# Chapter 8. p82: how long must the two pre-registered tests run?
# (1) Entry deletion, switchback by whole days: daily paid orders vary
#     from day to day with coefficient of variation cv; detect a
#     relative change delta in mean daily orders, two-sided alpha 0.05,
#     power 0.80: days per arm n = 2 (z_a + z_b)^2 cv^2 / delta^2.
# (2) Layer audit: request cycle time is log-normal with log-sd s; to
#     detect a ratio r of medians, requests per arm
#     n = 2 (z_a + z_b)^2 s^2 / (ln r)^2. Inputs illustrative.
import math
from statistics import NormalDist

Z = NormalDist().inv_cdf(0.975) + NormalDist().inv_cdf(0.80)
print("z_alpha + z_beta = %.4f" % Z)
print()
print("(1) days per arm for the entry switchback")
print("  delta     cv 0.10  cv 0.15  cv 0.20")
for delta in (0.05, 0.085, 0.10, 0.15):
    cells = [math.ceil(2 * Z * Z * cv * cv / delta ** 2)
             for cv in (0.10, 0.15, 0.20)]
    print("  %5.3f    %6d   %6d   %6d" % ((delta,) + tuple(cells)))
print()
print("(2) requests per arm for the layer audit")
print("  ratio    s 0.4   s 0.6   s 0.8")
for r in (1.10, 1.20, 1.27, 1.50):
    cells = [math.ceil(2 * Z * Z * s * s / math.log(r) ** 2)
             for s in (0.4, 0.6, 0.8)]
    print("  %4.2f    %5d   %5d   %5d" % ((r,) + tuple(cells)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《布雷斯悖论瓦德罗普均衡、购买入口加法反效与协同层级的无秩序代价母典》第 8 章「实证闭环:删入口的整时段切换实验、组织层级审计与预注册方案」(版本 1.0.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-246 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026246,
  author    = {Fan, Yafu},
  title     = {Braess's Paradox and Wardrop Equilibrium: When One More Purchase Entry or One More Coordination Layer Slows Everyone Down},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论